Re: [patch] epoll use a single inode ...

[Christoph Hellwig <hch@infradead.org>]

On Wed, Mar 07, 2007 at 08:16:14AM +0100, Eric Dumazet wrote:
> Crazy ideas : (some readers are going to kill me)
> 
> 1) Use the low order bit of f_path.dentry to say : this pointer is not a 
> pointer to a dentry but the inode pointer (with the low order bit set to 1)
> 
> OR
> 
> 2) file->f_path.dentry set to NULL for this special files (so that we dont 
> need to dput() and cache line ping pong the common dentry each time we 
> __fput() a pipe/socket.

No way on either one.  f_path.dentry always beeing there is an assumption
we make all over the place, and changing that would be a big regression
for code qualityand reliability all over the place.

Face it folks, memory is generally cheap, and we're not going to uglify
huge amounts of code to shave of a little bit.

[and that is only in reply to this one, the single dentry optimizations
 for epoll and friends are perfecltly fine from the high level view]

> Same trick could be used for file->f_path.mnt, because there is a big SMP 
> cache line ping/pong to maintain a mnt_count on pipe/sockets mountpoint 
> while these file systems cannot be un-mounted)

Same thing as above.  We might do a hack to not refcount these vfsmounts,
but we definitively want to keep the invariant of f_path.mnt never
beeing NULL.